The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install several well-known and reputable software tools, including plantuml (PyPI), mermaid-cli (NPM), and tldraw (NPM). These tools are industry-standard for the diagramming tasks described.
[INDIRECT_PROMPT_INJECTION]: The skill functions by interpreting user-supplied code (e.g., Python, JavaScript) or data (e.g., JSON, YAML) to generate visual diagrams. This represents a potential vulnerability surface where malicious content within the user's files could attempt to manipulate the agent.
Ingestion points: Ingestion of source code in the code-flow-visualizer skill and structured data in the json-data-visualizer skill.
Boundary markers: The skill does not explicitly instruct the agent to use boundary markers or to ignore instructions embedded in the input data.
Capability inventory: The skill itself is purely informational/documentation-based and does not execute system commands or perform network requests with user data.
Sanitization: Input data sanitization is not addressed in the guides.
[SAFE]: The skill correctly references and provides usage examples for well-known, established third-party services and APIs, such as kroki.io, plantuml.com, mermaid.live, and excalidraw.com, all of which are relevant to the skill's primary purpose.

diagram-skills